Method and apparatus for dual radar assisted detection of continuous blood pressure based on physiological guidance

Abstract

The application provides a method and device for continuously detecting blood pressure based on physiological guidance and dual radar assistance, and applies to the technical field of electromagnetic induction. The method comprises the following steps: acquiring a plurality of heart pulse signals of a target object, wherein the plurality of heart pulse signals represent mixed signals of a plurality of chest electromagnetic echo signals and a plurality of neck electromagnetic echo signals; determining a target signal from the plurality of heart pulse signals, wherein the target signal represents a heart pulse signal with a base frequency definition greater than a predetermined definition threshold; performing physiological feature extraction processing on the target signal to obtain physiological feature information of the target object; performing feature extraction on the physiological feature information and the target signal by using a continuous blood pressure prediction generation model to obtain a first feature and a second feature, wherein the first feature represents a four-dimensional feature corresponding to systolic pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic pressure; and performing feature fusion processing on the first feature and the second feature to obtain a target continuous blood pressure.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic induction technology, and more specifically to a method and apparatus for continuous blood pressure detection based on physiological guidance and dual radar assistance. Background Technology

[0002] Continuous blood pressure monitoring of targets is currently an important auxiliary means for early cardiovascular risk management. In existing methods, wearable sensors are typically used for continuous auxiliary monitoring of blood pressure; however, these sensors pose a risk of irritation to the target's skin. Therefore, non-contact radar systems can also be used for continuous auxiliary monitoring of blood pressure.

[0003] In the process of realizing the above-mentioned inventive concept, it was found through research that in the process of continuous auxiliary monitoring of the target's blood pressure using single-point non-contact radar in related technologies, problems such as physiological noise, signal separation and low measurement resolution make it difficult to monitor the target's blood pressure with high reliability. Furthermore, in the process of acquiring physiological information, it is difficult to apply the hemodynamic mechanism to the analysis process, resulting in technical problems that make it difficult to monitor the target's blood pressure with high accuracy. Summary of the Invention

[0004] In view of the above problems, the present invention provides a method and apparatus for continuous blood pressure detection based on physiological guidance dual radar.

[0005] According to a first aspect of the present invention, a method for continuous blood pressure detection based on physiological guidance using dual radar is provided, comprising: acquiring multiple cardiac pulse signals of a target object, wherein the multiple cardiac pulse signals represent a mixed signal of multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals; determining a target signal from the multiple cardiac pulse signals, wherein the target signal represents a cardiac pulse signal with a fundamental frequency clarity greater than a predetermined clarity threshold; performing physiological feature extraction processing on the target signal to obtain physiological feature information of the target object; using a continuous blood pressure prediction generation model to extract features from the physiological feature information and the target signal to obtain a first feature and a second feature, wherein the first feature represents a four-dimensional feature corresponding to systolic blood pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic blood pressure; performing feature fusion processing on the first feature and the second feature to obtain the target continuous blood pressure, wherein the target continuous blood pressure represents the continuous changes in systolic and diastolic blood pressure of the target object within a predetermined time period.

[0006] A second aspect of the present invention provides a device for continuous blood pressure detection based on physiological guidance using dual radar, comprising: an acquisition module for acquiring multiple cardiac pulse signals of a target object, wherein the multiple cardiac pulse signals represent a mixed signal of multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals; a determination module for determining a target signal from the multiple cardiac pulse signals, wherein the target signal represents a cardiac pulse signal with a fundamental frequency clarity greater than a predetermined clarity threshold; a first extraction module for performing physiological feature extraction processing on the target signal to obtain physiological feature information of the target object; a second extraction module for performing feature extraction on the physiological feature information and the target signal using a continuous blood pressure prediction generation model to obtain a first feature and a second feature, wherein the first feature represents a four-dimensional feature corresponding to systolic blood pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic blood pressure; and a fusion module for performing feature fusion processing on the first feature and the second feature to obtain the target continuous blood pressure, wherein the target continuous blood pressure represents the continuous changes in systolic and diastolic blood pressure of the target object within a predetermined time period.

[0007] A third aspect of the present invention provides an electronic device comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the method described above.

[0008] A fourth aspect of the present invention also provides a computer-readable storage medium having executable instructions stored thereon, which, when executed by a processor, cause the processor to perform the methods described above.

[0009] A fifth aspect of the invention also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0010] The method and apparatus for continuous blood pressure detection based on physiological guidance of the present invention first acquires multiple cardiac pulse signals containing multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals, and then extracts multiple target signals from the multiple cardiac pulse signals based on the fundamental frequency clarity of each cardiac pulse signal and a predetermined clarity threshold. This realizes the selection of high-quality target signals from multiple cardiac pulse signals from the original radar signals full of noise and interference by using the fundamental frequency clarity of each signal as a screening criterion through multiple rounds of algorithms, so as to extract more accurate physiological feature information from the high-quality target signals.

[0011] According to an embodiment of the present invention, the target signal is then subjected to physiological feature extraction processing to extract accurate beat interval sequence, pulse arrival time sequence, and respiratory signal related physiological feature information of the target object. The accurate physiological feature information and high-quality target signal are then input into a pre-trained continuous blood pressure prediction generation model. The continuous blood pressure prediction generation model performs multi-resolution feature extraction on the extracted accurate physiological feature information and high-quality target signal, thereby obtaining first features corresponding to systolic blood pressure and second features corresponding to diastolic blood pressure at different fine-grained levels. Upsampling and other feature fusion processing are performed on the first and second features of different dimensions to obtain fused features, which are then decoded to reconstruct the predicted continuous blood pressure of the target object. This achieves multi-dimensional feature capture and feature analysis extraction of chest and neck signals after interference noise processing using a dual-radar detection system. This study comprehensively analyzes blood pressure from multiple perspectives, including hemodynamics and cardiac dynamics. Based on accurate physiological characteristic information, it utilizes an optimized, robust continuous blood pressure prediction model. Multiple feature separations and fusions are then performed on the extracted physiological characteristic information to fully integrate and predict features corresponding to diastolic and systolic blood pressure. This yields a highly accurate and reliable target continuous blood pressure reading, which can assist professionals in subsequent processing, improving efficiency while saving costs.

[0012] According to embodiments of the present invention, furthermore, by conducting extensive model learning training on the continuous blood pressure prediction generation model, the trained continuous blood pressure prediction generation model can perform multiple rounds of feature extraction and fusion processing on the input physiological information and target signals. This allows for in-depth mining and analysis of the relationships between systolic blood pressure and neck and chest signals, as well as the relationships between diastolic blood pressure and respiratory sinus arrhythmia and changes in intrathoracic pressure, hidden within the physiological feature information and target signals. Thus, the target continuous blood pressure can be accurately generated based on features from different dimensions, resulting in a continuous blood pressure prediction generation model with high robustness, high reliability, high efficiency, and high output accuracy. Attached Figure Description

[0013] The above-described features, other objects, and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings, in which:

[0014] Figure 1 The diagram illustrates an application scenario of a physiologically guided dual-radar-assisted continuous blood pressure detection method according to an embodiment of the present invention.

[0015] Figure 2A flowchart of a method for continuous blood pressure detection based on physiological guidance dual radar assisted detection according to an embodiment of the present invention is shown;

[0016] Figure 3 A schematic diagram of a method for continuous blood pressure detection based on physiological guidance using dual radars according to an embodiment of the present invention is shown.

[0017] Figure 4 A schematic diagram illustrating the processing of physiological feature information and target signals using a continuous blood pressure prediction generation model according to an embodiment of the present invention is shown.

[0018] Figure 5a A schematic diagram showing a comparison between the results output by the method of the present invention according to the first embodiment of the present invention and actual continuous blood pressure is shown;

[0019] Figure 5b A schematic diagram showing a comparison between the results output by the method of the present invention according to the second embodiment of the present invention and actual continuous blood pressure is shown;

[0020] Figure 5c A schematic diagram showing a comparison between the results output by the method of the present invention according to the third embodiment of the present invention and actual continuous blood pressure is shown;

[0021] Figure 5d A schematic diagram showing a comparison between the results output by the method of the present invention according to the fourth embodiment of the present invention and actual continuous blood pressure is shown;

[0022] Figure 5e A schematic diagram showing a comparison between the results output by the method of the present invention according to the fifth embodiment of the present invention and actual continuous blood pressure is shown;

[0023] Figure 5f A schematic diagram showing a comparison between the results output by the method of the present invention according to the sixth embodiment of the present invention and actual continuous blood pressure is shown;

[0024] Figure 5g A schematic diagram showing a comparison between the results output by the method of the present invention according to the seventh embodiment of the present invention and actual continuous blood pressure is shown;

[0025] Figure 5h A schematic diagram showing a comparison between the results output by the method of the present invention according to the eighth embodiment of the present invention and actual continuous blood pressure is shown;

[0026] Figure 6 A structural block diagram of a device for continuous blood pressure detection based on physiological guidance using dual radars, according to an embodiment of the present invention, is shown.

[0027] Figure 7A block diagram of an electronic device for a physiologically guided dual-radar-assisted method for continuous blood pressure detection according to an embodiment of the present invention is shown. Detailed Implementation

[0028] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the invention. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of the invention for ease of explanation. However, it will be apparent that one or more embodiments may be practiced without these specific details. Furthermore, descriptions of well-known structures and techniques are omitted in the following description to avoid unnecessarily obscuring the concept of the invention.

[0029] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0030] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0031] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0032] In the technical solution of this invention, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0033] In recent years, continuous blood pressure monitoring of targets has become an important auxiliary means for early cardiovascular risk management. The ability to continuously monitor systolic blood pressure (SBP) and diastolic blood pressure (DBP) provides substantial benefits for guiding target-related matters, especially in assisting relevant personnel in assessing cardiovascular risk and making personalized medication decisions.

[0034] Related technologies can utilize wearable sensors for continuous auxiliary monitoring of a target's blood pressure. However, wearable sensors pose a risk of skin irritation, potentially causing problems beyond just blood pressure. Alternatively, non-contact radar can also be used for continuous auxiliary monitoring of a target's blood pressure. These methods primarily rely on single-channel peripheral waveform analysis, neglecting complex hemodynamic interactions. For example, compensatory changes in peripheral resistance in hypotensive patients can lead to waveform distortion, and the pulse signal acquired by radar is more susceptible to physiological noise due to its propagation characteristics. Multi-point peripheral arterial wave analysis is currently a method bridging non-contact sensing and clinical-grade hemodynamic monitoring. This analysis can estimate pulse transit time; as blood flow distance increases, the measured transit time becomes less sensitive to noise, allowing for wave velocity estimation. Furthermore, these methods struggle to fully utilize the rich physiological information in radar signals, often employing end-to-end predictive models that neglect hemodynamic mechanisms or require individual calibration of the estimated results. During the research and development process, it was found that in the process of using single-point non-contact radar to continuously assist in monitoring the blood pressure of a target, problems such as physiological noise, signal separation, and low measurement resolution make it difficult to monitor the blood pressure of the target with high reliability. Furthermore, in the process of acquiring physiological information, it is difficult to apply hemodynamic mechanisms to the analysis process, which leads to technical problems that make it difficult to monitor the blood pressure of the target with high accuracy.

[0035] In view of this, embodiments of the present invention provide a method for continuous blood pressure detection based on physiological guidance using dual radar, comprising: acquiring multiple cardiac pulse signals of a target object, wherein the multiple cardiac pulse signals represent a mixed signal of multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals; determining a target signal from the multiple cardiac pulse signals, wherein the target signal represents a cardiac pulse signal with a fundamental frequency clarity greater than a predetermined clarity threshold; performing physiological feature extraction processing on the target signal to obtain physiological feature information of the target object; using a continuous blood pressure prediction generation model to extract features from the physiological feature information and the target signal to obtain a first feature and a second feature, wherein the first feature represents a four-dimensional feature corresponding to systolic blood pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic blood pressure; performing feature fusion processing on the first feature and the second feature to obtain the target continuous blood pressure, wherein the target continuous blood pressure represents the continuous changes in systolic and diastolic blood pressure of the target object within a predetermined time period.

[0036] Figure 1 The diagram illustrates an application scenario of a physiologically guided dual-radar-assisted method for continuous blood pressure monitoring according to an embodiment of the present invention.

[0037] like Figure 1 As shown, the application scenario according to this embodiment may include a first radar device 101, a second radar device 102, a target object 103, and a receiver 104. The first radar device 101 and the second radar device 102 are used to send electromagnetic wave signals to the target object 103.

[0038] Users can use the first radar device 101 and the second radar device 102 to interact with the target object 103 and the receiver 104 to receive or send signals, etc.

[0039] Receiver 104 can be a receiver that receives various electromagnetic echo signals, such as receiving and processing electromagnetic echo signals sent by the first radar device 101 and the second radar device 102 (for example only). Receiver 104 can analyze and process the received electromagnetic echo signals and other data, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0040] It should be noted that the physiologically guided dual-radar assisted continuous blood pressure detection method provided in this embodiment of the invention can generally be executed by receiver 104. Correspondingly, the physiologically guided dual-radar assisted continuous blood pressure detection device provided in this embodiment of the invention can generally be housed in receiver 104. The physiologically guided dual-radar assisted continuous blood pressure detection method provided in this embodiment of the invention can also be executed by a receiver or receiver cluster that is different from receiver 104 and capable of communicating with the first radar device 101, the second radar device 102, and / or receiver 104. Correspondingly, the physiologically guided dual-radar assisted continuous blood pressure detection device provided in this embodiment of the invention can also be housed in a receiver or receiver cluster that is different from receiver 104 and capable of communicating with the first radar device 101, the second radar device 102, and / or receiver 104.

[0041] It should be understood that Figure 1 The number of the first radar device, second radar device, target object, and receiver shown is merely illustrative. Any number of radar devices, target objects, and receivers can be included depending on the implementation requirements.

[0042] The following will be based on Figure 1 The described scene, through Figures 2-5h The method for continuous blood pressure detection based on physiological guidance using dual radars according to embodiments of the present invention will be described in detail.

[0043] Figure 2 A flowchart of a method for continuous blood pressure detection based on physiological guidance using dual radars, according to an embodiment of the present invention, is shown.

[0044] like Figure 2 As shown, the method for continuous blood pressure detection based on physiological guidance using dual radars in this embodiment includes operations S210 to S250.

[0045] In operation S210, multiple cardiac pulse signals of the target object are acquired.

[0046] According to an embodiment of the present invention, multiple cardiac pulse signals characterize a mixed signal of multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals.

[0047] According to an embodiment of the present invention, a radar is placed at predetermined positions at the chest and neck regions of a target object, respectively. The transmitting antennas of the two radars transmit radar signals to the chest and neck regions of the target object, respectively. Electromagnetic echo signals are generated in the chest and neck regions and returned to the receiving antennas of the two radars. The millimeter-wave radar may have multiple transmitting and receiving antennas. The predetermined distance between the radars can be 40 cm, and the relevant parameters of the two radars can be set to a starting frequency of 77 GHz and a sweep slope of 99.987. Frequency sweep duration 40 Frequency sweep repetition rate 200Hz, idle time 80 The analog-to-digital converter (ADC) has 256 samples, an ADC sampling rate of 8000kbps, and a receive gain of 48dB.

[0048] According to an embodiment of the present invention, multiple cardiac pulse signals are reflected back to multiple receiving antennas from different spatial points on the target object through different transmission channels. These multiple target signals may include multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals. The chest electromagnetic echo signal can reflect relevant activity information of the target object's cardiac region and may be a cardiac signal. The neck electromagnetic echo signal can reflect the vibration information of the target object's carotid artery and may be a vibration caused by pressure wave transmission.

[0049] In operation S220, the target signal is determined from multiple cardiac pulse signals.

[0050] According to an embodiment of the present invention, the target signal represents a cardiac pulse signal with a fundamental frequency clarity greater than a predetermined clarity threshold.

[0051] According to embodiments of the present invention, the target signal can be a signal extracted from multiple cardiac pulse signals by using fundamental frequency clarity as the signal extraction standard. Furthermore, a predetermined clarity threshold or a predetermined number can be set as a reference for the number of high-quality target signals selected, wherein there can be multiple target signals.

[0052] In operation S230, physiological feature extraction processing is performed on the target signal to obtain the physiological feature information of the target object.

[0053] According to embodiments of the present invention, physiological characteristic information may include physiological information related to the target object, such as the beat interval sequence, pulse arrival time sequence, and respiratory signals.

[0054] According to an embodiment of the present invention, by acquiring the current physiological information of the target object, a multi-dimensional comprehensive analysis of the aforementioned physiological characteristics can be performed to facilitate the prediction and generation of the target object's continuous blood pressure.

[0055] In operation S240, a continuous blood pressure prediction generation model is used to extract features from physiological characteristic information and target signals to obtain the first feature and the second feature.

[0056] According to an embodiment of the present invention, the first feature represents a four-dimensional feature corresponding to systolic blood pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic blood pressure.

[0057] According to embodiments of the present invention, the continuous blood pressure prediction generation model can be pre-trained using a sample training dataset. During application, the model can be directly used to perform multi-resolution feature extraction on the physiological characteristic information of the input target object, obtaining a first feature and a second feature corresponding to diastolic and systolic blood pressure, respectively. The first feature can be a fine-grained, low-level feature, representing the temporal relationship between neck and chest signals related to systolic blood pressure. The second feature can be a coarse-grained, high-level feature, representing respiratory sinus arrhythmias and intrathoracic pressure changes that affect diastolic blood pressure.

[0058] In operation S250, the first feature and the second feature are fused to obtain the target continuous blood pressure.

[0059] According to an embodiment of the present invention, the target continuous blood pressure characterizes the continuous changes in systolic and diastolic blood pressure of a target subject within a predetermined time period.

[0060] According to an embodiment of the present invention, after performing multi-resolution feature extraction processing on physiological feature information and target signal using a continuous blood pressure prediction generation model, the extracted first feature and second feature are fused, and then the fused feature is further decoded to obtain the target continuous blood pressure containing continuous diastolic and systolic blood pressure.

[0061] According to an embodiment of the present invention, the target continuous blood pressure can be used as auxiliary information for predicting the target object. It should be noted that the target continuous blood pressure obtained here is only as reference information for medical staff to make diagnoses and is not used as a direct diagnostic result.

[0062] For example, two radars are used to transmit electromagnetic signals to the chest and neck regions of a target object, respectively. Multiple electromagnetic echo signals are preliminarily processed to obtain multiple cardiac pulse signals. The target signal is then identified from these cardiac pulse signals based on fundamental frequency clarity. Physiological feature extraction processing is then performed on the target signal and the multiple cardiac pulse signals to obtain the target object's physiological feature information. A continuous blood pressure prediction generation model is then used to perform multi-resolution feature extraction processing on the physiological feature information to obtain a first feature related to systolic blood pressure and a second feature related to diastolic blood pressure. The first and second features are then fused to obtain and generate a target continuous blood pressure for predicting the target object. This allows scientists or professionals in related fields to monitor the relevant status of the subject based on the target continuous blood pressure and make appropriate judgments.

[0063] According to an embodiment of the present invention, multiple cardiac pulse signals, including multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals, are first acquired. Then, based on the fundamental frequency clarity of each cardiac pulse signal and a predetermined clarity threshold, multiple target signals are extracted from the multiple cardiac pulse signals. This achieves the goal of selecting high-quality target signals from multiple cardiac pulse signals from noisy and interference-filled raw radar signals by using the fundamental frequency clarity of each signal as a screening criterion through multiple rounds of algorithms, so as to extract more accurate physiological feature information from the high-quality target signals.

[0064] According to an embodiment of the present invention, the target signal is then subjected to physiological feature extraction processing to extract accurate beat interval sequence, pulse arrival time sequence, and respiratory signal related physiological feature information of the target object. The accurate physiological feature information and high-quality target signal are then input into a pre-trained continuous blood pressure prediction generation model. The continuous blood pressure prediction generation model performs multi-resolution feature extraction on the extracted accurate physiological feature information and high-quality target signal, thereby obtaining first features corresponding to systolic blood pressure and second features corresponding to diastolic blood pressure at different fine-grained levels. Upsampling and other feature fusion processing are performed on the first and second features of different dimensions to obtain fused features, which are then decoded to reconstruct the predicted continuous blood pressure of the target object. This achieves multi-dimensional feature capture and feature analysis extraction of chest and neck signals after interference noise processing using a dual-radar detection system. This study comprehensively analyzes blood pressure from multiple perspectives, including hemodynamics and cardiac dynamics. Based on accurate physiological characteristic information, it utilizes an optimized, robust continuous blood pressure prediction model. Multiple feature separations and fusions are then performed on the extracted physiological characteristic information to fully integrate and predict features corresponding to diastolic and systolic blood pressure. This yields a highly accurate and reliable target continuous blood pressure reading, which can assist professionals in subsequent processing, improving efficiency while saving costs.

[0065] According to embodiments of the present invention, furthermore, by conducting extensive model learning training on the continuous blood pressure prediction generation model, the trained continuous blood pressure prediction generation model can perform multiple rounds of feature extraction and fusion processing on the input physiological information and target signals. This allows for in-depth mining and analysis of the relationships between systolic blood pressure and neck and chest signals, as well as the relationships between diastolic blood pressure and respiratory sinus arrhythmia and changes in intrathoracic pressure, hidden within the physiological feature information and target signals. Thus, the target continuous blood pressure can be accurately generated based on features from different dimensions, resulting in a continuous blood pressure prediction generation model with high robustness, high reliability, high efficiency, and high output accuracy.

[0066] It should be noted that the target continuous blood pressure obtained by the present invention is only an intermediate result and cannot be used to directly derive a diagnosis or health status from the target continuous blood pressure obtained by the method according to the present invention.

[0067] According to embodiments of the present invention, a method for acquiring multiple cardiac pulse signals of a target object may include the following operations.

[0068] According to an embodiment of the present invention, multiple initial electromagnetic echo signals of the target object are acquired.

[0069] According to an embodiment of the present invention, the plurality of initial electromagnetic echo signals include an initial mixed signal of a plurality of chest initial electromagnetic echo signals reflected by the chest region of the target object and a plurality of neck initial electromagnetic echo signals reflected by the neck region.

[0070] According to an embodiment of the present invention, electromagnetic signals are first transmitted to the target object using two radars located in the chest region and the neck region, and then multiple initial electromagnetic echo signals of the chest and multiple initial electromagnetic echo signals of the neck are received from various spatial points and radar channels using a receiving antenna array.

[0071] According to an embodiment of the present invention, the initial electromagnetic echo signal is subjected to arctangent demodulation processing to obtain multiple phase signals.

[0072] According to an embodiment of the present invention, the phase signal can be obtained as expressed by formula (1).

[0073] (1);

[0074] in, It can be represented as a phase signal, Q(t) can be represented as the real part of the initial electromagnetic echo signal, and I(t) can be represented as the imaginary part of the initial electromagnetic echo signal.

[0075] According to an embodiment of the present invention, multiple phase signals are subjected to second-order differential filtering to obtain multiple cardiac pulse signals.

[0076] According to an embodiment of the present invention, multiple initial electromagnetic echo signals are subjected to noise and interference suppression processing using a second-order differential filter that can robustly suppress signal noise, thereby obtaining multiple cardiac pulse signals, which can be expressed according to formula (2).

[0077] (2);

[0078] Among them, V m (t) can be represented as the cardiac pulse signal, N can be represented as the half-width length of the second-order differential filter, h can be represented as the sampling time step, and s can be represented as the second-order differential filter.

[0079] According to an embodiment of the present invention, the length of the second-order differential filter used above can be as shown in formula (3). The coefficients of the second-order differential filter can be calculated by recursion. The coefficients of the second-order differential filter need to satisfy the symmetry relationship and boundary conditions. The coefficients of the second-order differential filter can be as shown in formula (4), the symmetry relationship can be as shown in formula (5), and the boundary conditions can be as shown in formula (6).

[0080] L=2N+1(3)

[0081] Here, L can be represented as the length of the second-order differential filter.

[0082] (4);

[0083] Here, s[k] can be represented as the coefficients of the second-order differential filter, and k can be represented as the index.

[0084] s[L-1-k]= s[k](5;

[0085] Here, s[L-1-k] can be characterized as the symmetric coefficients of the second-order differential filter coefficients s[k].

[0086] s[N]=1(6;

[0087] Here, s[N] can be characterized as the boundary condition.

[0088] According to an embodiment of the present invention, the length L of the second-order differential filter can be set to 43, and can also be specifically set according to the specific application scenario.

[0089] According to an embodiment of the present invention, multiple initial electromagnetic echo signals reflected from the chest region and the neck region of the target object are first acquired. Then, phase extraction is performed on the multiple initial electromagnetic echo signals to obtain multiple phase signals. Then, second-order differential filtering is performed on the multiple phase signals to obtain multiple cardiac pulse signals. This achieves preliminary filtering of multiple initial electromagnetic echo signals to suppress and filter out some noise and interference, improve the quality of the signals to be processed subsequently, and avoid noise interference affecting the processing process.

[0090] According to embodiments of the present invention, a method for determining a target signal from multiple cardiac pulse signals may include the following operations.

[0091] According to an embodiment of the present invention, multiple pre-processed cardiac pulse signals are correlated by a Harmonic-Conscious Synchrosqueezed Cardiac Tracker (HCSCT) algorithm, thereby extracting accurate physiological feature information of the target object from the multiple cardiac pulse signals. The Harmonic-Conscious Synchrosqueezed Cardiac Tracker algorithm can be implemented based on synchronous compressed wavelet transform, harmonic verification, and determination of fundamental frequency clarity information based on the fundamental frequency of the target heart rate.

[0092] According to an embodiment of the present invention, multiple cardiac pulse signals are subjected to synchronous compressed wavelet transform and harmonic verification processing to obtain the target heart rate fundamental frequency of the target object.

[0093] According to an embodiment of the present invention, by performing synchronous compressed wavelet transform and harmonic verification processing on multiple cardiac pulse signals, the global cardiac pulse frequency that can be applied globally, i.e. the target cardiac pulse frequency, can be determined from the cardiac pulse frequency corresponding to each cardiac pulse signal.

[0094] According to an embodiment of the present invention, the fundamental frequency clarity information of each cardiac pulse signal is determined based on the fundamental frequency clarity selection rule and the target heart rate fundamental frequency.

[0095] According to an embodiment of the present invention, using a determined target heart rate fundamental frequency that can be adapted globally, the fundamental frequency clarity of each cardiac pulse signal is calculated based on the fundamental frequency clarity selection rule, thereby obtaining the fundamental frequency clarity information of each cardiac pulse signal. The fundamental frequency clarity information of each cardiac pulse signal is used as a selection index and sorted from largest to smallest to facilitate the selection of multiple target signals from multiple cardiac pulse signals.

[0096] According to an embodiment of the present invention, a target signal is determined from multiple cardiac pulse signals based on a predetermined sharpness threshold and the fundamental frequency sharpness information of each cardiac pulse signal.

[0097] According to embodiments of the present invention, after sorting based on the fundamental frequency clarity information of each cardiac pulse signal, a predetermined clarity threshold can be used to select cardiac pulse signals with a resolution greater than the predetermined clarity threshold as multiple target signals. Simultaneously, a predetermined number can be set as a selection constraint for the target signals. From the multiple cardiac pulse signals, a predetermined number of cardiac pulse signals with higher fundamental frequency clarity can be selected as multiple target signals, and a corresponding set of target signals can be constructed. .

[0098] According to an embodiment of the present invention, by performing synchronous compressed wavelet transform and harmonic verification processing on multiple cardiac pulse signals, the target heart rate fundamental frequency of the target object is obtained. Then, based on the fundamental frequency clarity selection rule and the target heart rate fundamental frequency, the fundamental frequency clarity information of each cardiac pulse signal is determined. Then, according to a predetermined clarity threshold and the fundamental frequency clarity information of each cardiac pulse signal, the target signal is determined from multiple cardiac pulse signals. This realizes the processing of multiple cardiac pulse signals by the harmonic sensing synchronous compressed wavelet heart tracker algorithm. The fundamental frequency clarity of each cardiac pulse signal can be used as a selection metric to select multiple high-quality and relevant target signals from multiple cardiac pulse signals, and the target fundamental frequency is determined at the same time, so as to extract the current accurate physiological feature information of the target object from multiple target signals.

[0099] According to an embodiment of the present invention, a method for obtaining the target heart rate fundamental frequency of a target object by performing synchronous compressed wavelet transform and harmonic verification processing on multiple cardiac pulse signals may include the following operations.

[0100] According to an embodiment of the present invention, the mother wavelet function is used to perform continuous wavelet transform processing on multiple cardiac pulse signals to obtain wavelet coefficients corresponding to each cardiac pulse signal.

[0101] According to an embodiment of the present invention, by performing continuous wavelet transform (CWT) processing on multiple cardiac pulse signals, the time-frequency characteristics of each cardiac pulse signal can be analyzed to obtain a matrix of wavelet coefficients that can reflect the time-frequency characteristics.

[0102] According to an embodiment of the present invention, before performing continuous wavelet transform processing on multiple cardiac pulse signals, it is necessary to separate multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals from the multiple cardiac pulse signals, so as to perform continuous wavelet transform processing on the multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals respectively. Taking the continuous wavelet transform processing on the multiple chest electromagnetic echo signals as an example, the wavelet coefficients of the chest electromagnetic echo signals can be expressed as shown in formula (7).

[0103] (7);

[0104] Among them, W c (a, b) can be represented as wavelet coefficients of the chest electromagnetic echo signal, where a can be represented as a scale parameter and b as a time shift parameter. c (t) can be characterized as the wavelet coefficients of the chest electromagnetic echo signal. It can be represented as the mother wavelet function. It can be characterized as the conjugate of the mother wavelet function.

[0105] According to an embodiment of the present invention, multiple neck electromagnetic echo signals are also subjected to continuous wavelet transform processing in the same manner, thereby obtaining the wavelet coefficients of each neck electromagnetic echo signal.

[0106] According to an embodiment of the present invention, the wavelet coefficients corresponding to each cardiac pulse signal are subjected to synchronous compression transformation to obtain multiple target time-frequency matrices.

[0107] According to an embodiment of the present invention, in order to further improve the energy concentration and resolution of the time-frequency characteristics, the wavelet coefficients of each chest electromagnetic echo signal and each neck electromagnetic echo signal can be processed by Synchrosqueezing Transform (SST). Through the Synchrosqueezing Transform, the energy of the wavelet coefficients can be remapped to their true instantaneous frequency, thereby effectively concentrating the energy related to the heartbeat while suppressing noise and interference components.

[0108] According to an embodiment of the present invention, taking the synchronous compression transformation of wavelet coefficients of multiple chest electromagnetic echo signals as an example, the target time-frequency matrix of the chest electromagnetic echo signal can be represented by formula (8).

[0109] (8);

[0110] in, This can be represented as the target time-frequency matrix of the chest electromagnetic echo signal. It can be characterized as frequency. It can be characterized as the instantaneous frequency of the chest electromagnetic echo signal extracted from the phase information of the wavelet coefficients obtained after wavelet transform of the chest electromagnetic echo signal. It can be represented as the Dirac function.

[0111] According to an embodiment of the present invention, the wavelet coefficients of multiple neck electromagnetic echo signals are also subjected to synchronous compression transformation in the same manner, thereby obtaining the target time-frequency matrix of each neck electromagnetic echo signal.

[0112] According to an embodiment of the present invention, harmonic verification processing is performed on multiple target time-frequency matrices to obtain the target heart rate fundamental frequency of the target object.

[0113] According to an embodiment of the present invention, harmonic verification processing is performed on the target time-frequency matrix of each chest electromagnetic echo signal and the target time-frequency matrix of each neck electromagnetic echo signal, thereby ensuring the consistency of each element in the subsequently extracted beat interval sequence at spatial points, and fundamental frequency detection is performed to obtain the target heart rate fundamental frequency of the target object.

[0114] According to an embodiment of the present invention, a method for obtaining the target heart rate fundamental frequency of a target object by performing harmonic verification processing on multiple target time-frequency matrices may include the following operations.

[0115] According to an embodiment of the present invention, based on the time shift parameter, the value of each element in multiple target time-frequency matrices is accumulated to obtain the cardiac energy spectrum corresponding to each target time-frequency matrix.

[0116] According to an embodiment of the present invention, by accumulating the energy values ​​at all time points at predetermined frequencies in the target time-frequency matrix of each chest electromagnetic echo signal and the target time-frequency matrix of each neck electromagnetic echo signal based on time shift parameters, the cardiac energy spectrum corresponding to the target time-frequency matrix of each chest electromagnetic echo signal and the cardiac energy spectrum corresponding to the target time-frequency matrix of each neck electromagnetic echo signal can be obtained.

[0117] According to an embodiment of the present invention, taking the energy accumulation processing of the target time-frequency matrix of multiple chest electromagnetic echo signals as an example, the cardiac energy spectrum of each chest electromagnetic echo signal can be expressed by formula (9).

[0118] (9);

[0119] in, It can be characterized as the cardiac energy spectrum of the chest electromagnetic echo signal.

[0120] According to an embodiment of the present invention, the target time-frequency matrices of multiple neck electromagnetic echo signals are also subjected to energy accumulation processing in the same manner, thereby obtaining the cardiac energy spectrum of each neck electromagnetic echo signal.

[0121] According to an embodiment of the present invention, multiple cardiac energy spectra are smoothed using a filter to obtain multiple target energy spectra.

[0122] According to an embodiment of the present invention, after obtaining the cardiac energy spectrum of each chest electromagnetic echo signal and the cardiac energy spectrum of each neck electromagnetic echo signal, each cardiac energy spectrum is smoothed again using a filter to reduce its noise fluctuations and obtain the target energy spectrum.

[0123] According to an embodiment of the present invention, based on the frequency peak selection rule, the target heart rate fundamental frequency of the target object is determined from the first cardiac frequency interval corresponding to multiple target energy spectra.

[0124] According to an embodiment of the present invention, the frequency peak selection rule can be characterized as selecting the two highest energy peaks from the target energy spectrum.

[0125] According to an embodiment of the present invention, a method for determining the target heart rate fundamental frequency of a target object from a first cardiac frequency interval corresponding to multiple target energy spectra based on a frequency peak selection rule may include the following operations.

[0126] According to an embodiment of the present invention, based on the frequency peak selection rule, a first cardiac peak and a second cardiac peak are determined from the first cardiac frequency interval corresponding to each target energy spectrum.

[0127] According to embodiments of the present invention, the first cardiac frequency range can typically be set between 0.6 Hz and 2.5 Hz, but is not limited thereto; the first cardiac frequency range can also be specifically set according to actual circumstances. First cardiac peak value. It can be characterized as the first peak with the highest energy selected based on the frequency peak selection rule, and the second cardiac peak. It can be characterized as the second highest energy peak selected based on the frequency peak selection rule.

[0128] For example, based on the frequency peak selection rule, the two highest energy peaks are selected in the range of 0.6Hz to 2.5Hz in each energy spectrum. The first selected peak is taken as the first heartbeat peak, and the second selected peak is taken as the second heartbeat peak.

[0129] According to an embodiment of the present invention, the harmonic relationship of the frequency of the first cardiac peak and the frequency of the second cardiac peak of each target energy spectrum is verified to determine the intermediate heart rate fundamental frequency of each target energy spectrum.

[0130] According to an embodiment of the present invention, harmonic relationship verification can be characterized as verifying whether a harmonic relationship is satisfied between the frequencies of the selected first and second cardiac peaks. If the frequency of the second cardiac peak is an integer multiple of the frequency of the first cardiac peak, the harmonic relationship is satisfied; if the frequency of the second cardiac peak is not an integer multiple of the frequency of the first cardiac peak, the harmonic relationship is not satisfied. When the harmonic relationship is satisfied, the intermediate heart rate fundamental frequency corresponding to the target energy spectrum is determined as the first cardiac peak and its frequency. When the harmonic relationship is not satisfied, the frequency of the peak with the highest energy between the first and second cardiac peaks is selected as the intermediate heart rate fundamental frequency corresponding to the target energy spectrum. For example, when the harmonic relationship is not satisfied, the frequency of the first cardiac peak is 1.7 Hz with an energy of 30 J, and the frequency of the second cardiac peak is 0.8 Hz with an energy of 20 J, thus the frequency of the first cardiac peak, 1.7 Hz, is taken as the intermediate heart rate fundamental frequency corresponding to the target energy spectrum.

[0131] According to an embodiment of the present invention, the above-mentioned harmonic relationship verification and determination of the intermediate heart rate fundamental frequency are performed on each target energy spectrum to obtain the intermediate heart rate fundamental frequency of each target energy spectrum.

[0132] According to an embodiment of the present invention, a target heart rate base frequency is determined from multiple intermediate heart rate base frequencies based on a heart rate base frequency selection rule.

[0133] According to an embodiment of the present invention, the heart rate fundamental frequency selection rule can be characterized as selecting the two intermediate heart rate fundamental frequencies with the highest occurrence frequency from a plurality of intermediate heart rate fundamental frequencies and verifying the harmonic relationship again. When the harmonic relationship is satisfied, the first intermediate heart rate fundamental frequency is taken as the target heart rate fundamental frequency. When the harmonic relationship is not satisfied, the fundamental frequency with the higher corresponding energy among the two intermediate heart rate fundamental frequencies is taken as the target heart rate fundamental frequency. For example, there are 10 intermediate heart rate fundamental frequencies. Among them, the intermediate heart rate fundamental frequency with a frequency of 10Hz and the intermediate heart rate fundamental frequency with a frequency of 0.8Hz appear most frequently, 5 times and 3 times respectively. The energy of the intermediate heart rate fundamental frequency with a frequency of 1.7Hz is 20J, and the energy of the intermediate heart rate fundamental frequency with a frequency of 0.8Hz is 10J. The intermediate heart rate fundamental frequency with a frequency of 1.7Hz is taken as the first potential heart rate fundamental frequency, and the intermediate heart rate fundamental frequency with a frequency of 0.8Hz is taken as the second potential heart rate fundamental frequency. The harmonic relationship between the first potential heart rate fundamental frequency and the second potential heart rate fundamental frequency is verified. The second potential heart rate fundamental frequency is not an integer multiple of the first potential heart rate fundamental frequency, confirming that the harmonic relationship is not satisfied. The first potential heart rate fundamental frequency with the highest energy is taken as the target heart rate fundamental frequency.

[0134] According to an embodiment of the present invention, multiple wavelet coefficients representing the time-frequency characteristics of the signals are obtained by performing continuous wavelet transform on multiple cardiac pulse signals using a mother wavelet function. Then, a synchronous compression transform is performed on each wavelet coefficient, achieving energy concentration and improved resolution of the time-frequency characteristics. The synchronous compression transform remaps the energy of the wavelet coefficients to their true instantaneous frequencies, effectively concentrating the energy related to heartbeats while suppressing noise and interference components, thus facilitating the determination of the target heart rate fundamental frequency based on the obtained target time-frequency matrix. Then, based on time shift parameters, the energy of all time points in the target time-frequency matrix is ​​accumulated at a predetermined frequency to construct the cardiac energy spectrum of each cardiac pulse signal. Finally, after smoothing each cardiac energy spectrum by filtering out noise and interference, multiple target energy spectra are obtained. Based on the frequency peak selection rules, the heart rate fundamental frequency selection principle, and the corresponding cardiac frequency intervals that are compatible with the energy spectrum and intermediate heart rate fundamental frequencies, harmonic verification is performed to determine the target heart rate fundamental frequency. Harmonic characteristics are used to ensure that the heart rate intervals inferred from different spatial points have inherent physiological consistency. At the same time, the fundamental frequency clarity information corresponding to each cardiac pulse signal is calculated based on the corresponding peak energy. The fundamental frequency clarity information is used as a reference index to screen target signals from multiple cardiac pulse signals, and high-quality target signals are screened to facilitate the extraction of physiological feature information based on the target signals, thereby obtaining accurate physiological feature information with physiological consistency and high accuracy.

[0135] According to embodiments of the present invention, a method for determining the fundamental frequency clarity information of each cardiac pulse signal based on fundamental frequency clarity selection rules and target heart rate fundamental frequency may include the following operations.

[0136] According to an embodiment of the present invention, based on the fundamental frequency clarity selection rule, the third cardiac peak and the fourth cardiac peak are determined from the second cardiac frequency interval corresponding to each cardiac energy spectrum.

[0137] According to an embodiment of the present invention, the second cardiac frequency range is obtained based on the target heart rate base frequency and the first error base frequency.

[0138] According to an embodiment of the present invention, the fundamental frequency clarity selection rule can be characterized as selecting the highest energy peak from the cardiac energy spectrum, and locating the nearest minimum energy point to the right of the highest energy peak (in the direction of higher frequency) as the fourth cardiac peak.

[0139] According to an embodiment of the present invention, the first error base frequency can be a predetermined value. The second cardiac frequency range can be obtained by taking the target heart rate base frequency ± the first error base frequency. For example, if the first error base frequency is 0.4Hz and the target heart rate base frequency is 2Hz, then the second cardiac frequency range is 1.6Hz~2.4Hz.

[0140] For example, based on the fundamental frequency clarity selection rule, within the range of ±0.4Hz of the target heart rate fundamental frequency of each cardiac energy spectrum, the highest energy peak is selected as the third cardiac peak. Then, to the right of the third cardiac peak, the smallest energy point closest to the third cardiac peak is located as the fourth cardiac peak.

[0141] According to an embodiment of the present invention, the fundamental frequency clarity information of each cardiac pulse signal is obtained based on the energy information of the third cardiac peak and the energy information of the fourth cardiac peak of each cardiac energy spectrum.

[0142] According to an embodiment of the present invention, the fundamental frequency clarity information is obtained based on the ratio between the energy information of the third cardiac peak and the energy information of the fourth cardiac peak. The fundamental frequency clarity information can be as shown in formula (10).

[0143] (10);

[0144] Among them, R i This can be characterized as fundamental frequency resolution information, E peak，i The energy information that can be represented as the third cardiac peak, E min，i This can be characterized as the energy information of the fourth cardiac peak.

[0145] According to an embodiment of the present invention, by performing the above-described operation on each cardiac energy spectrum, the fundamental frequency clarity information of each cardiac energy spectrum can be obtained.

[0146] According to an embodiment of the present invention, by using a fundamental frequency clarity selection rule, a third cardiac peak and a fourth cardiac peak are determined from the second cardiac frequency interval corresponding to each cardiac energy spectrum. Based on the energy information of the third cardiac peak and the fourth cardiac peak of each cardiac energy spectrum, the fundamental frequency clarity information of each cardiac pulse signal is confirmed. Thus, the fundamental frequency clarity information of each signal can be used as a screening index to select a portion of target signals with higher quality from multiple cardiac pulse signals, so as to prepare for the extraction of accurate physiological feature information.

[0147] According to an embodiment of the present invention, the target signal can be P target signals, where P is a positive integer, and the physiological feature information can include the target chest beat interval sequence and the target neck beat interval sequence, the pulse arrival time sequence and the respiratory signal.

[0148] According to embodiments of the present invention, a method for extracting physiological features from a target signal to obtain physiological feature information of a target object may include the following operations.

[0149] According to an embodiment of the present invention, for the p-th target signal, a plurality of candidate peaks are determined from the p-th target signal based on a first cardiac frequency period.

[0150] According to an embodiment of the present invention, p = 1, 2, 3, ..., P, and both p and P are positive integers.

[0151] According to an embodiment of the present invention, the first cardiac frequency period can be used to divide each target signal. By dividing the target signal, a candidate peak can be determined within each first cardiac frequency period, thereby obtaining a series of candidate peaks for the p-th target signal, avoiding an insufficient number of extracted candidate peak samples. The first cardiac frequency period can be set according to specific circumstances.

[0152] According to an embodiment of the present invention, multiple cardiac time differences are calculated based on the times of each adjacent candidate peak among multiple candidate peaks.

[0153] For example, there are currently 5 candidate peaks. The cardiac time difference between the 1st and 2nd candidate peaks is 500ms, the cardiac time difference between the 2nd and 3rd candidate peaks is 1000ms, the cardiac time difference between the 3rd and 4th candidate peaks is 700ms, and the cardiac time difference between the 4th and 5th candidate peaks is 600ms.

[0154] According to an embodiment of the present invention, multiple cardiac time differences are sorted to obtain the first beat-by-beat interval sequence of the p-th target signal.

[0155] According to an embodiment of the present invention, multiple cardiac parallaxes are sorted in chronological order to obtain the first beat interval sequence of the p-th target signal. For example, the first beat interval sequence is represented as 500ms, 1000ms, 700ms, and 600ms.

[0156] According to an embodiment of the present invention, based on the signal source, the first beat interval sequence is divided to obtain a first chest beat interval sequence or a first neck beat interval sequence.

[0157] According to an embodiment of the present invention, the signal source can be characterized as the region of reflected electromagnetic echo signal, specifically the chest region and the neck region.

[0158] According to an embodiment of the present invention, the above-described operation is performed on each target signal to obtain a first beat interval sequence for each target signal. Then, the multiple target signals are divided according to the chest region and the neck region. The target signals corresponding to the chest are divided into one group, and their first beat interval sequence is determined as the first chest beat interval sequence. The target signals corresponding to the neck are divided into another group, and their first beat interval sequence is determined as the first neck beat interval sequence. Thus, the two groups of signals are analyzed separately without interference.

[0159] According to an embodiment of the present invention, a target chest beat interval sequence is determined from a plurality of first chest beat interval sequences based on a predetermined cardiac cycle.

[0160] According to an embodiment of the present invention, the difference between each cardiac time difference in the target chest beat interval sequence and the predetermined cardiac cycle is smaller than the other cardiac time differences in each first chest beat interval sequence corresponding to each cardiac time difference.

[0161] According to an embodiment of the present invention, both the target chest beat interval sequence and the target neck beat interval sequence may include multiple cardiac time differences. The process of determining the optimal target chest beat interval sequence from multiple first chest beat interval sequences based on a predetermined cardiac cycle and the process of determining the optimal target neck beat interval sequence from multiple first neck beat interval sequences can both be as shown in formula (11).

[0162] (11);

[0163] in, It can be represented as the optimal cardiac time difference selected from multiple first chest beat interval sequences, where i can be represented as the index of the i-th cardiac time difference in multiple first chest beat interval sequences, and M can be represented as the total number of cardiac time differences in multiple first chest beat interval sequences. This can be represented as a calculation to find the minimum value among M cardiac time differences. This can be represented as the i-th cardiac time difference. This can be characterized as the target heart rate fundamental frequency. It can be characterized as a predetermined cardiac cycle.

[0164] For example, there are currently three first chest beat interval sequences corresponding to the chest: the first first chest beat interval sequence is 500ms, 1000ms, 700ms, 900ms; the second first chest beat interval sequence is 600ms, 100ms, 300ms, 500ms; and the third first chest beat interval sequence is 1200ms, 1400ms, 600ms, 550ms, with a predetermined heart rate cycle of 600ms.

[0165] The cardiac time difference of the first column in the three first chest beat interval sequences was calculated with the absolute value difference of the predetermined cardiac cycle to obtain 100ms, 0ms, and 600ms. Based on the principle of minimum value, the first cardiac time difference in the second first chest beat interval sequence was determined as the first cardiac time difference in the target chest beat interval sequence.

[0166] The absolute difference between the second cardiac time difference in the three first chest beat interval sequences and the predetermined cardiac cycle was calculated to obtain 400ms, 500ms, and 800ms. Based on the principle of minimum value, the second cardiac time difference in the first first chest beat interval sequence was determined as the second cardiac time difference in the target chest beat interval sequence.

[0167] The absolute difference between the cardiac time difference in the third column of the three first chest beat interval sequences and the predetermined cardiac cycle was calculated to obtain 100ms, 300ms, and 200ms. Based on the principle of minimum value, the third cardiac time difference in the first first chest beat interval sequence was determined as the third cardiac time difference in the target chest beat interval sequence.

[0168] The cardiac time difference of the fourth column in the three first chest beat interval sequences was calculated by comparing the absolute value difference with the predetermined cardiac cycle, yielding values ​​of 300ms, 100ms, and 50ms. Based on the principle of minimizing the value, the fourth cardiac time difference in the third first chest beat interval sequence was determined as the fourth cardiac time difference in the target chest beat interval sequence. Thus, the target chest beat interval sequence was constructed.

[0169] According to an embodiment of the present invention, a target neck beat interval sequence is determined from a plurality of first neck beat interval sequences based on a predetermined cardiac cycle.

[0170] According to an embodiment of the present invention, the difference between each cardiac time difference in the target neck beat interval sequence and the predetermined cardiac cycle is smaller than the other cardiac time differences in each first neck beat interval sequence corresponding to each cardiac time difference.

[0171] According to an embodiment of the present invention, the process of determining the target neck beat interval sequence is consistent with the process of determining the target chest beat interval sequence, and will not be described again here.

[0172] According to embodiments of the present invention, the target signal corresponding to the chest and the target signal corresponding to the neck with the highest fundamental frequency clarity information can also be directly selected from multiple target signals corresponding to the chest and multiple target signals corresponding to the neck based on the fundamental frequency clarity information. These two signals are then directly used as the chest beat interval signal and the neck beat interval signal. Then, multiple candidate peaks are determined from the chest beat interval signal and the neck beat interval signal, and multiple cardiac time differences of the chest beat interval signal and the neck beat interval signal are calculated. The time series composed of the multiple cardiac time differences of the chest beat interval signal is used as the target chest beat interval sequence, and the time series composed of the multiple cardiac time differences of the neck beat interval signal is used as the target neck beat interval sequence.

[0173] According to an embodiment of the present invention, a pulse arrival time sequence is obtained based on the difference between the target chest pulse interval sequence and the target neck pulse interval sequence.

[0174] According to an embodiment of the present invention, the pulse arrival time sequence can be obtained as expressed by formula (12).

[0175] (12);

[0176] Wherein, PAT(t) can be characterized as the pulse arrival time series, IBI c (t) can be characterized as the target chest beat-by-beat interval sequence, IBI n (t) can be characterized as the target neck beat interval sequence.

[0177] According to an embodiment of the present invention, the phase signals of multiple initial electromagnetic echo signals are subjected to low-pass filtering to obtain a breathing signal.

[0178] According to an embodiment of the present invention, the initial electromagnetic echo signal can be characterized as the signal received by the radar receiving antenna array without any processing.

[0179] According to an embodiment of the present invention, the respiratory signal can be obtained as expressed by formula (13).

[0180] (13);

[0181] Among them, Rm (t) can be represented as a breathing signal, and F(·) can be represented as a low-pass filter.

[0182] According to embodiments of the present invention, by using a low-pass filter to perform low-pass filtering on the phase signals of multiple initial electromagnetic echo signals, high-frequency cardiac impulse components and other noises can be effectively filtered out, separating relevant low-frequency signals that reflect respiratory motion. The respiratory signals may include respiratory signals corresponding to the chest region and respiratory signals corresponding to the neck region.

[0183] According to an embodiment of the present invention, multiple candidate peaks are determined from each target signal, and the cardiac time difference is calculated based on each pair of adjacent candidate peaks to construct multiple first beat interval sequences. Based on the signal source, the multiple first beat interval sequences are divided into corresponding chest and neck regions to obtain multiple first chest beat interval sequences and multiple first neck beat interval sequences. Then, based on a predetermined cardiac cycle, a target chest beat interval sequence is determined from the multiple first chest beat interval sequences, and a target neck beat interval sequence is determined from the multiple first neck beat interval sequences. This enables the acquisition of more accurate and stable beat interval sequences through a frequency-constrained peak selection method. Furthermore, based on the difference between the target chest beat interval sequence and the target neck beat interval sequence, a pulse arrival time sequence is obtained. The phase signal of the initial electromagnetic echo signal is low-pass filtered to effectively filter out high-frequency cardiac pulsation components and other noise in the signal, separating the low-frequency signal that reflects respiratory motion, i.e., the respiratory signal. This enables the accurate extraction of physiological feature information from the target signal, so that the physiological feature information can be output to the model for continuous blood pressure prediction and generation.

[0184] According to an embodiment of the present invention, the method for extracting features from physiological feature information and target signals using a continuous blood pressure prediction generation model to obtain a first feature and a second feature may include the following operations.

[0185] According to an embodiment of the present invention, a sequence to be predicted is generated based on physiological characteristic information and target signal.

[0186] According to an embodiment of the present invention, the target signal may include multiple target chest electromagnetic echo signals and multiple target neck electromagnetic echo signals. The target chest beat interval sequence, the target neck beat interval sequence, the pulse arrival time sequence, the respiratory signal, the target chest electromagnetic echo signals, and the target neck electromagnetic echo signals are input into a continuous blood pressure prediction generation model. The continuous blood pressure prediction generation model generates a prediction sequence for continuous blood pressure based on the above sequences and signals. The prediction sequence can be a multivariate time series, allowing the continuous blood pressure prediction generation model to obtain continuous diastolic blood pressure and continuous systolic blood pressure based on continuous features in the time series.

[0187] According to an embodiment of the present invention, initial features are extracted from the sequence to be predicted to obtain initial features.

[0188] According to an embodiment of the present invention, the initial feature characterizes a mixture of features corresponding to diastolic and systolic blood pressure.

[0189] According to an embodiment of the present invention, multi-resolution feature extraction is performed on the initial features to obtain the first feature and the second feature.

[0190] According to embodiments of the present invention, multi-level feature extraction can be performed on the initial features. Specifically, low-level (fine-grained) feature extraction can be performed on the initial features to obtain low-dimensional features, which can be used to identify the temporal relationship between neck and chest signals related to systolic blood pressure. High-level (coarse-grained) feature extraction can be performed on the initial features to obtain high-dimensional features, which can be used to model respiratory sinus arrhythmias and intrathoracic pressure changes that affect diastolic blood pressure.

[0191] According to an embodiment of the present invention, the specific sequence to be predicted, initial features, first features, and second features can be expressed and extracted with reference to the formulas used in the training process.

[0192] According to an embodiment of the present invention, a sequence to be predicted is generated based on physiological feature information and target signal. Initial feature extraction is performed on the sequence to be predicted to obtain mixed features corresponding to diastolic and systolic blood pressure. Multi-resolution feature extraction is performed on the initial features to obtain first and second features. This realizes multi-level and multi-granular feature extraction from physiological feature information and target signal, thereby mining and analyzing features related to systolic and diastolic blood pressure from multiple dimensions and levels, so as to generate a target continuous blood pressure containing continuous diastolic and systolic blood pressure information based on the time features of the sequence to be predicted.

[0193] According to an embodiment of the present invention, after extracting the first feature and the second feature from the initial features, the first feature and the second feature can be encoded using a continuous blood pressure prediction generation model to obtain the corresponding first encoded feature and the second encoded feature. Then, the first encoded feature and the second encoded feature can be fused to obtain the target fused feature. The target fused feature is then decoded to directly predict the target continuous blood pressure of the target object.

[0194] According to an embodiment of the present invention, the continuous blood pressure prediction generative model is trained in the following manner, and the training method may include the following operations.

[0195] According to an embodiment of the present invention, an initial model to be trained and a sample training dataset are obtained.

[0196] According to an embodiment of the present invention, the sample training dataset includes multiple consecutive blood pressure samples, multiple training physiological feature information, multiple training chest signals, and multiple training neck signals.

[0197] According to embodiments of the present invention, the multiple training physiological characteristic information may include multiple training chest beat interval sequences, multiple training neck beat interval sequences, multiple training pulse arrival time sequences, and multiple training respiratory signals.

[0198] According to an embodiment of the present invention, each continuous blood pressure sample includes continuous diastolic blood pressure information and continuous systolic blood pressure information.

[0199] According to an embodiment of the present invention, continuous blood pressure samples can be as shown in formula (14).

[0200] (14);

[0201] Wherein, y(t) can be represented as a continuous blood pressure sample, SBP(t) can be represented as continuous sample systolic blood pressure information, and DBP(t) can be represented as continuous sample diastolic blood pressure information.

[0202] According to an embodiment of the present invention, multiple training physiological feature information, multiple training chest signals, and multiple training neck signals are input into the initial model to be trained for sequence generation processing to obtain multiple training sequences.

[0203] According to an embodiment of the present invention, the training sequence can be as shown in formula (15).

[0204] (15);

[0205] Here, x(t) can be represented as the training sequence, and T can be represented as the time length of the sequence. It can be represented as the real number field, Rc (t) can be represented as the training breathing signal corresponding to the chest region. The training sequence can be a 6-dimensional feature sequence, V c (t) can be represented as the training chest signal, V n (t) can be characterized as the training neck signal.

[0206] According to an embodiment of the present invention, feature extraction is performed on multiple training sequences to obtain multiple first training features and multiple second training features.

[0207] According to an embodiment of the present invention, the first training feature represents a four-dimensional training feature corresponding to systolic blood pressure, and the second training feature represents an eight-dimensional training feature corresponding to diastolic blood pressure.

[0208] According to an embodiment of the present invention, initial features are first extracted from multiple training sequences to obtain initial training features. These initial training features can characterize the mixed training features corresponding to diastolic and systolic blood pressure.

[0209] According to an embodiment of the present invention, the initial features for training can be as shown in formula (16).

[0210] (16);

[0211] Among them, v r It can be represented as the initial features for training, F r (·) can be characterized as the process of extracting initial training features from the training sequence, and D can be characterized as the feature dimension.

[0212] According to an embodiment of the present invention, after the initial training features are extracted, multi-resolution feature extraction is performed on the initial training features to obtain the first training features and the second training features.

[0213] According to embodiments of the present invention, multi-level feature extraction can be performed on the initial training features. Specifically, low-level (fine-grained) feature extraction can be performed on the initial training features to obtain low-dimensional features, which can be used to train the identification of the temporal relationship between neck and chest signals related to systolic blood pressure. High-level (coarse-grained) feature extraction can be performed on the initial training features to obtain high-dimensional features, which can be used to train the modeling of respiratory sinus arrhythmias and intrathoracic pressure changes that affect diastolic blood pressure.

[0214] According to an embodiment of the present invention, the first training feature may be as shown in formula (17), and the second training feature may be as shown in formula (18).

[0215] (17);

[0216] Among them, v lThis can be represented as the first training feature, F l (·) can be characterized as the process of extracting the first training feature from the initial training features.

[0217] (18);

[0218] Among them, v h This can be represented as the second training feature, F h (·) can be characterized as the process of extracting a second training feature from the initial training features.

[0219] According to an embodiment of the present invention, a plurality of first training features and a plurality of second training features are input into an encoder for encoding processing to obtain a plurality of first training encoded features and a plurality of second training encoded features.

[0220] According to an embodiment of the present invention, the first training coding feature can be as shown in formula (19), and the second training coding feature can be as shown in formula (20).

[0221] (19);

[0222] Among them, z l This can be represented as the first training encoded feature, G l (·) can be represented as the process of encoding the first training feature.

[0223] (20);

[0224] Among them, z h This can be represented as the second training encoding feature, G h (·) can be represented as the process of encoding the second training feature.

[0225] According to an embodiment of the present invention, multiple first training coding features and multiple second training coding features are subjected to feature fusion processing, and the resulting multiple training fusion features are input to a decoder for decoding processing to obtain multiple predicted continuous blood pressures.

[0226] According to an embodiment of the present invention, the second training coding feature is first upsampled to improve its resolution, and then the feature dimensions of the upsampled second training coding feature and the first training coding feature are aligned to achieve feature fusion and obtain training fusion features.

[0227] According to an embodiment of the present invention, the training fusion features can be expressed as formula (21).

[0228] (twenty one);

[0229] Among them, zo It can be represented as training fusion features, Concat(·) can be represented as alignment processing, and Upsample(·) can be represented as upsampling processing.

[0230] According to embodiments of the present invention, the above-described progressive upsampling method can preserve physiologically relevant details in signals from two radar sources, thereby enabling continuous prediction of diastolic and systolic blood pressure.

[0231] According to an embodiment of the present invention, after obtaining the training fusion features, the training fusion features can be decoded using a decoder to restore the predicted continuous blood pressure.

[0232] According to an embodiment of the present invention, continuous blood pressure can be predicted as shown in formula (22).

[0233] (twenty two);

[0234] in, It can be characterized as a predictor of continuous blood pressure.

[0235] According to an embodiment of the present invention, in practical applications, this step can directly output the target continuous blood pressure.

[0236] According to an embodiment of the present invention, a training loss value is calculated based on a loss function, according to multiple predicted continuous blood pressures and multiple continuous blood pressure samples.

[0237] According to an embodiment of the present invention, the training loss value can be calculated as shown in formula (23).

[0238] (twenty three);

[0239] Here, l can be represented as the training loss value.

[0240] According to an embodiment of the present invention, the model parameters of the initial model to be trained are adjusted based on the training loss value to obtain a trained continuous blood pressure prediction generation model.

[0241] According to an embodiment of the present invention, a predetermined number of training rounds and a predetermined loss threshold can be set. When the model's iterative training rounds meet the predetermined number of training rounds or the model's training loss value is less than the predetermined loss threshold, the training of the model can be stopped, and a trained continuous blood pressure prediction generation model can be obtained.

[0242] According to an embodiment of the present invention, multiple training physiological feature information, multiple training chest signals, and multiple training neck signals from the sample training dataset are input into the initial model to be trained. The initial model generates multiple training sequences based on these signals. Then, initial features are extracted from the training sequences to obtain multiple initial training features. Each initial training feature is then subjected to multi-resolution feature extraction to obtain multiple first training features and multiple second training features. These first and second training features are then specifically encoded. Finally, the encoded first and second training features are fused and decoded. The process involves processing the data to obtain multiple predicted continuous blood pressure readings generated during the training of the initial model to be trained. Then, based on the loss function, a training loss value is calculated from these multiple predicted continuous blood pressure readings and samples. The model parameters of the initial model to be trained are then adjusted based on a predetermined number of rounds or a predetermined loss threshold, resulting in a trained continuous blood pressure prediction generation model. This process achieves the training of the initial model. By utilizing a large amount of training data for comprehensive and repeated learning and training of the model's features, including feature extraction, encoding, fusion, and decoding, the robustness and maturity of the continuous blood pressure prediction generation model can be improved, outputting highly accurate target continuous blood pressure readings to adapt to blood pressure prediction generation in various environments and improve efficiency.

[0243] Figure 3 A schematic diagram of a method for continuous blood pressure detection based on physiological guidance using dual radars, according to an embodiment of the present invention, is shown.

[0244] like Figure 3 As shown, two radars are used to transmit radar signals to the chest and neck regions of the target object and receive multiple initial electromagnetic echo signals. The multiple initial electromagnetic echo signals are processed by second-order differential filtering to obtain multiple cardiac pulse signals. Then, the multiple cardiac pulse signals are processed by harmonic sensing synchronous compressed wavelet transform to determine the target heart rate fundamental frequency and multiple target signals. Then, based on the target heart rate fundamental frequency, the target beat interval sequence, pulse arrival time sequence and respiratory signal are extracted from the multiple target signals and multiple initial electromagnetic echo signals. The target beat interval sequence, pulse arrival time sequence and respiratory signal are then input into the continuous blood pressure prediction generation model for multi-resolution feature extraction, encoding, feature fusion and decoding processing to output the target continuous blood pressure.

[0245] Figure 4 A schematic diagram illustrating the processing of physiological feature information and target signals using a continuous blood pressure prediction generation model according to an embodiment of the present invention is shown.

[0246] like Figure 4As shown, the continuous blood pressure prediction generation model performs multi-resolution feature extraction on the input sequence to be predicted, which consists of physiological feature information, target chest electromagnetic echo signal, and target neck electromagnetic echo signal, to obtain the first feature and the second feature. Then, the first feature and the second feature are encoded to obtain the first encoded feature and the second encoded feature. The second encoded feature is then upsampled and then aligned with the first encoded feature and fused to obtain the target fused feature. The target fused feature is then decoded to obtain the target continuous blood pressure containing continuous diastolic blood pressure and continuous systolic blood pressure.

[0247] Figure 5a A schematic diagram showing a comparison between the results output by the method of the present invention according to the first embodiment of the present invention and actual continuous blood pressure is shown.

[0248] like Figure 5a As shown, the blue and purple curves represent the predicted systolic and diastolic blood pressure values ​​of the present invention, while the orange and yellow curves represent the true systolic and diastolic blood pressure values ​​of the prior art. Testing was performed on multiple target subjects. Compared to actual continuous blood pressure, the method of the present invention shows a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP), with a correlation coefficient r = 0.725. The mean of the predicted systolic blood pressure values ​​is also relatively small. The standard deviation is 2.73. The error distribution shows a slightly higher predicted systolic blood pressure (DBP) value overall than the true value (r=0.740). The error between the predicted and true values ​​for continuous diastolic blood pressure (DBP) is small, with a correlation coefficient (r=0.740) for the predicted values. The standard deviation is -0.27. The error ratio is 1.51, indicating that the overall predicted diastolic blood pressure is approximately consistent with the true value in the error distribution. Specifically, the mean error of systolic / diastolic blood pressure in this invention is -0.33 / -0.17 mmHg, and the mean absolute error is 7.32 / 5.69 mmHg. According to existing evaluation criteria, the dual-radar system of this invention achieves 6.43 mmHg (systolic blood pressure) and 5.32 mmHg (diastolic blood pressure) in terms of mean absolute deviation, meeting the key clinical benchmark of below 7 mmHg.

[0249] Figure 5b A schematic diagram showing a comparison between the results output by the method of the present invention according to the second embodiment of the present invention and actual continuous blood pressure is shown.

[0250] like Figure 5bAs shown, the blue and purple curves represent the predicted systolic and diastolic blood pressure values ​​of the present invention, while the orange and yellow curves represent the true systolic and diastolic blood pressure values ​​of the prior art. The target object was tested at different detection positions (lying down). Compared to the actual continuous blood pressure, the method of the present invention has a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP), with a correlation coefficient r = 0.922 between the predicted and true values. The mean of the predicted systolic blood pressure is... The standard deviation is 1.24. The error distribution shows that the predicted systolic blood pressure is slightly higher than the true value, with a value of 1.65. The predicted diastolic blood pressure (DBP) has a smaller error compared to the true value, with a correlation coefficient (r) of 0.732. The mean of the predicted diastolic blood pressure is... The standard deviation is -0.14. The value is 1.55, indicating that the overall predicted diastolic blood pressure is slightly higher than the true value in the error distribution.

[0251] Figure 5c A schematic diagram showing a comparison between the results output by the method of the present invention according to the third embodiment of the present invention and actual continuous blood pressure is shown.

[0252] like Figure 5c As shown, the blue and purple curves represent the predicted systolic and diastolic blood pressure values ​​of this invention, while the orange and yellow curves represent the true systolic and diastolic blood pressure values ​​of the prior art. Multiple target subjects were tested in different seasons. Compared to actual continuous blood pressure, the method of this invention shows a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP), with a correlation coefficient r = 0.769. The mean of the predicted systolic blood pressure values ​​is also relatively small. The standard deviation is 1.53. The error distribution shows a slightly higher predicted systolic blood pressure (DBP) value overall than the true value (4.19). The error between the predicted and true values ​​for continuous diastolic blood pressure (DBP) is small, with a correlation coefficient (r) of 0.627. The mean of the predicted diastolic blood pressure is... The standard deviation is -0.64. The value is 1.92, and the overall predicted value of diastolic blood pressure in the error distribution is approximately consistent with the true value.

[0253] Figure 5d A schematic diagram showing a comparison between the results output by the method of the present invention according to the fourth embodiment of the present invention and actual continuous blood pressure is shown.

[0254] like Figure 5dAs shown, the blue and purple curves represent the predicted systolic and diastolic blood pressure values ​​of this invention, while the orange and yellow curves represent the true systolic and diastolic blood pressure values ​​of the prior art. By using a new detection radar to detect multiple targets, the continuous blood pressure output by this invention shows a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP) compared to the actual continuous blood pressure. The correlation between the predicted and true values ​​is r=0.811, and the mean of the predicted systolic blood pressure is... The standard deviation is 1.95. The error distribution shows that the predicted systolic blood pressure is slightly higher than the true value, with a correlation coefficient of 1.79. The predicted diastolic blood pressure (DBP) has a smaller error compared to the true value, with a correlation coefficient r = 0.799. The mean of the predicted diastolic blood pressure is... The standard deviation is -0.24. The error value is 1.62, indicating that the overall predicted diastolic blood pressure is slightly lower than the true value in the error distribution.

[0255] Figure 5e A schematic diagram showing a comparison between the results output by the method of the present invention according to the fifth embodiment of the present invention and actual continuous blood pressure is shown. Figure 5f A schematic diagram showing a comparison between the results output by the method of the present invention according to the sixth embodiment of the present invention and actual continuous blood pressure is shown. Figure 5g A schematic diagram showing a comparison between the results output by the method of the present invention according to the seventh embodiment of the present invention and actual continuous blood pressure is shown.

[0256] like Figures 5e to 5g As shown, the blue and purple curves represent the predicted systolic and diastolic blood pressure values ​​of the present invention, while the orange and yellow curves represent the true systolic and diastolic blood pressure values ​​of the prior art. Figure 5e In order to conduct testing on multiple target individuals in the morning, Figure 5f In order to conduct testing on multiple target individuals in the afternoon, Figure 5g To conduct testing on multiple target individuals at night.

[0257] Figure 5e The present invention outputs a target continuous blood pressure reading that, compared to the actual continuous blood pressure, shows a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP). The correlation between the predicted and true values ​​is r=0.695, and the mean of the predicted systolic blood pressure is... The standard deviation is -0.53. The error distribution shows that the predicted systolic blood pressure is slightly lower than the true value, with a value of 3.97. The error between the predicted and true values ​​for continuous diastolic blood pressure (DBP) is small, with a correlation coefficient r = 0.637. The mean of the predicted diastolic blood pressure is... The standard deviation is 0.54. The value is 1.85, indicating that the overall predicted diastolic blood pressure is slightly higher than the true value in the error distribution.

[0258] Figure 5f Compared to the actual continuous blood pressure, the target continuous blood pressure output by this invention shows a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP). The correlation between the predicted and true values ​​is r=0.809, the mean μ of the predicted systolic blood pressure is 1.70, and the standard deviation is... The error distribution shows that the predicted systolic blood pressure is slightly higher than the true value, with a value of 3.17. The error between the predicted and true values ​​for continuous diastolic blood pressure (DBP) is small, with a correlation coefficient (r) of 0.835. The mean μ of the predicted diastolic blood pressure is -0.52, and the standard deviation is... The value is 2.76, indicating that the overall predicted diastolic blood pressure is slightly lower than the true value in the error distribution.

[0259] Figure 5g The present invention outputs a target continuous blood pressure reading that, compared to the actual continuous blood pressure, shows a smaller error between the predicted and true values ​​of continuous systolic blood pressure (SBP). The correlation between the predicted and true values ​​is r=0.870, and the mean of the predicted systolic blood pressure is... The standard deviation is 1.85. The error distribution shows that the predicted systolic blood pressure is slightly higher than the true value, with a value of 5.09. The error between the predicted and true values ​​for continuous diastolic blood pressure (DBP) is small, with a correlation coefficient r = 0.697. The mean of the predicted diastolic blood pressure is... The standard deviation is 3.61. The value is 2.57, indicating that the overall predicted diastolic blood pressure is slightly higher than the true value in the error distribution.

[0260] Figure 5h A schematic diagram showing a comparison between the results output by the method of the present invention according to the eighth embodiment of the present invention and actual continuous blood pressure is shown.

[0261] like Figure 5h As shown, the blue and purple curves represent the predicted systolic and diastolic blood pressure values ​​of this invention, while the orange and yellow curves represent the true systolic and diastolic blood pressure values ​​of the prior art. The target continuous blood pressure output by this invention was compared with the actual continuous blood pressure in the morning and evening of the same day. It can be seen that in the two morning blood pressure measurements, the errors between the predicted and true values ​​of continuous systolic blood pressure (SBP) and continuous diastolic blood pressure (DBP) are smaller, with a correlation coefficient r=0.661 for systolic blood pressure and r=0.625 for diastolic blood pressure. In the two afternoon blood pressure measurements, the errors between the predicted and true values ​​of continuous systolic blood pressure (SBP) and continuous diastolic blood pressure (DBP) are smaller, with a correlation coefficient r=0.839 for systolic blood pressure and r=0.619 for diastolic blood pressure.

[0262] Figure 6 A structural block diagram of a device for continuous blood pressure detection based on physiological guidance using dual radars, according to an embodiment of the present invention, is shown.

[0263] like Figure 6 As shown, the device for continuous blood pressure detection based on physiological guidance dual radar in this embodiment includes: an acquisition module 610, a determination module 620, a first extraction module 630, a second extraction module 640, and a fusion module 650.

[0264] The acquisition module 610 is used to acquire multiple cardiac pulse signals of the target object, wherein the multiple cardiac pulse signals represent a mixed signal of multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals. The acquisition module 610 can be used to perform the operation S210 described above, which will not be repeated here.

[0265] The determination module 620 is used to determine a target signal from multiple cardiac pulse signals, wherein the target signal represents a cardiac pulse signal with a fundamental frequency clarity greater than a predetermined clarity threshold. The determination module 620 can be used to perform the operation S220 described above, which will not be repeated here.

[0266] The first extraction module 630 is used to perform physiological feature extraction processing on the target signal to obtain the physiological feature information of the target object. The first extraction module 630 can be used to execute the operation S230 described above, which will not be repeated here.

[0267] The second extraction module 640 is used to extract features from physiological feature information and target signals using a continuous blood pressure prediction generation model, obtaining a first feature and a second feature. The first feature represents a four-dimensional feature corresponding to systolic blood pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic blood pressure. The second extraction module 640 can be used to execute the operation S240 described above, which will not be repeated here.

[0268] The fusion module 650 is used to perform feature fusion processing on the first feature and the second feature to obtain the target continuous blood pressure, wherein the target continuous blood pressure represents the continuous changes in systolic and diastolic blood pressure of the target object within a predetermined time period. The fusion module 650 can be used to perform the operation S250 described above, which will not be repeated here.

[0269] According to embodiments of the present invention, any plurality of modules among the acquisition module 610, determination module 620, first extraction module 630, second extraction module 640, and fusion module 650 may be combined into one module, or any one of these modules may be split into multiple modules. Alternatively, at least part of the functionality of one or more of these modules may be combined with at least part of the functionality of other modules and implemented in one module. According to embodiments of the present invention, at least one of the acquisition module 610, determination module 620, first extraction module 630, second extraction module 640, and fusion module 650 may be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in hardware or firmware, or in any one of software, hardware, and firmware implementations, or in a suitable combination of any of these. Alternatively, at least one of the acquisition module 610, determination module 620, first extraction module 630, second extraction module 640 and fusion module 650 may be at least partially implemented as a computer program module, which can perform corresponding functions when the computer program module is run.

[0270] Figure 7 A block diagram of an electronic device for a physiologically guided dual-radar-assisted method for continuous blood pressure detection according to an embodiment of the present invention is shown.

[0271] like Figure 7 As shown, an electronic device according to an embodiment of the present invention includes a processor 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage portion 708 into a random access memory (RAM) 703. The processor 701 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 701 may also include onboard memory for caching purposes. The processor 701 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present invention.

[0272] RAM 703 stores various programs and data required for the operation of the electronic device. Processor 701, ROM 702, and RAM 703 are interconnected via bus 704. Processor 701 executes various operations of the method flow according to embodiments of the present invention by executing programs in ROM 702 and / or RAM 703. It should be noted that the programs may also be stored in one or more memories other than ROM 702 and RAM 703. Processor 701 may also execute various operations of the method flow according to embodiments of the present invention by executing programs stored in said one or more memories.

[0273] According to embodiments of the present invention, the electronic device may further include an input / output (I / O) interface 705, which is also connected to a bus 704. The electronic device may also include one or more of the following components connected to the I / O interface 705: an input section 706 including a keyboard, mouse, etc.; an output section 707 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 708 including a hard disk, etc.; and a communication section 709 including a network interface card such as a LAN card, modem, etc. The communication section 709 performs communication processing via a network such as the Internet. A drive 710 is also connected to the I / O interface 705 as needed. A removable medium 711, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 710 as needed so that computer programs read from it can be installed into the storage section 708 as needed.

[0274] The present invention also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of the present invention.

[0275] According to embodiments of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of the present invention, a computer-readable storage medium may include ROM 702 and / or RAM 703 and / or one or more memories other than ROM 702 and RAM 703 described above.

[0276] Embodiments of the present invention also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code enables the computer system to implement the physiologically guided dual-radar-assisted continuous blood pressure detection method provided in the embodiments of the present invention.

[0277] When the computer program is executed by the processor 701, it performs the functions defined in the system / apparatus of this invention. According to embodiments of the invention, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0278] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 709, and / or installed from a removable medium 711. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0279] In such an embodiment, the computer program can be downloaded and installed from a network via the communication section 709, and / or installed from the removable medium 711. When the computer program is executed by the processor 701, it performs the functions defined in the system of this embodiment of the invention. According to embodiments of the invention, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0280] Those skilled in the art will understand that the features described in the various embodiments and / or claims of the present invention can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in the present invention. In particular, the features described in the various embodiments and / or claims of the present invention can be combined or combined in various ways without departing from the spirit and teachings of the present invention. All such combinations and / or combinations fall within the scope of the present invention.

Claims

1. A device for continuous blood pressure detection based on physiological guidance and dual radar assistance, characterized in that, include: The acquisition module is used to acquire multiple cardiac pulse signals of the target object, wherein the multiple cardiac pulse signals represent a mixed signal of multiple chest electromagnetic echo signals and multiple neck electromagnetic echo signals; A determining module is configured to determine a target signal from the plurality of cardiac pulse signals. Determining the target signal includes: based on a fundamental frequency clarity selection rule and a target heart rate fundamental frequency, determining the highest energy value from the second cardiac frequency interval corresponding to each cardiac energy spectrum as a third cardiac peak, and determining the lowest energy value located to the right of the third cardiac peak and closest to it as a fourth cardiac peak; obtaining the fundamental frequency clarity information of each cardiac pulse signal based on the energy information of the third and fourth cardiac peaks in each cardiac energy spectrum; and determining the cardiac pulse signal with a fundamental frequency clarity greater than the predetermined clarity threshold from the plurality of cardiac pulse signals as the target signal based on a predetermined clarity threshold and the fundamental frequency clarity information of each cardiac pulse signal. The first extraction module is used to perform physiological feature extraction processing on the target signal to obtain the physiological feature information of the target object; The second extraction module is used to extract features from the physiological feature information and the target signal using a continuous blood pressure prediction generation model to obtain a first feature and a second feature. The first feature represents a four-dimensional feature corresponding to systolic blood pressure, and the second feature represents an eight-dimensional feature corresponding to diastolic blood pressure. The fusion module is used to perform feature fusion processing on the first feature and the second feature to obtain the target continuous blood pressure, which represents the continuous change of the systolic blood pressure and the diastolic blood pressure of the target object within a predetermined time period.

2. The apparatus according to claim 1, characterized in that, The determining module is further configured to: perform synchronous compressed wavelet transform and harmonic verification processing on the multiple cardiac pulse signals to obtain the target heart rate fundamental frequency of the target object.

3. The apparatus according to claim 2, characterized in that, The determining module is also used for: Using the mother wavelet function, continuous wavelet transform processing is performed on the multiple cardiac pulse signals to obtain wavelet coefficients corresponding to each cardiac pulse signal; The wavelet coefficients corresponding to each cardiac pulse signal are subjected to synchronous compression transformation to obtain multiple target time-frequency matrices. The harmonic verification process is performed on the multiple target time-frequency matrices to obtain the target heart rate fundamental frequency of the target object.

4. The apparatus according to claim 3, characterized in that, The determining module is also used for: Based on the time shift parameter, the value of each element in the multiple target time-frequency matrices is accumulated to obtain the cardiac energy spectrum corresponding to each target time-frequency matrix; Multiple cardiac energy spectra were smoothed using filters to obtain multiple target energy spectra; Based on the frequency peak selection rule, the target heart rate base frequency of the target object is determined from the first cardiac frequency interval corresponding to the multiple target energy spectra.

5. The apparatus according to claim 4, characterized in that, The determining module is also used for: Based on the frequency peak selection rule, the first cardiac peak and the second cardiac peak are determined from the first cardiac frequency interval corresponding to each target energy spectrum; The harmonic relationship of the frequencies of the first and second cardiac peaks of each target energy spectrum is verified to determine the intermediate heart rate fundamental frequency of each target energy spectrum; Based on the heart rate base frequency selection rule, the target heart rate base frequency is determined from multiple intermediate heart rate base frequencies.

6. The apparatus according to claim 1, characterized in that, The second cardiac frequency range is obtained based on the target heart rate base frequency and the first error base frequency.

7. The apparatus according to claim 1, characterized in that, The target signal includes P signals, and the physiological feature information includes the target chest beat interval sequence, the target neck beat interval sequence, the pulse arrival time sequence, and the respiratory signal. The first extraction module is also used for: For the p-th target signal, based on the first cardiac frequency cycle, multiple candidate peaks are determined from the p-th target signal, where p = 1, 2, 3, ..., P, and p and P are both positive integers; Based on the time of each adjacent candidate peak among the multiple candidate peaks, multiple cardiac time differences are calculated; The multiple cardiac time differences are sorted to obtain the first beat-by-beat interval sequence of the p-th target signal; Based on the signal source, the first beat interval sequence is divided to obtain a first chest beat interval sequence or a first neck beat interval sequence; Based on a predetermined cardiac cycle, the target chest beat interval sequence is determined from a plurality of first chest beat interval sequences, wherein the difference between each cardiac time difference in the target chest beat interval sequence and the predetermined cardiac cycle is less than the other cardiac time differences in each first chest beat interval sequence corresponding to each cardiac time difference. Based on the predetermined cardiac cycle, the target neck beat interval sequence is determined from a plurality of first neck beat interval sequences, wherein the difference between each cardiac time difference in the target neck beat interval sequence and the predetermined cardiac cycle is less than the other cardiac time differences in each first neck beat interval sequence corresponding to each cardiac time difference.

8. The apparatus according to claim 7, characterized in that, The first extraction module is also used for: The pulse arrival time sequence is obtained based on the difference between the target chest pulse interval sequence and the target neck pulse interval sequence; The breathing signal is obtained by low-pass filtering the phase signals of multiple initial electromagnetic echo signals.

9. The apparatus according to claim 1, characterized in that, The second extraction module is also used for: Based on the physiological characteristics and the target signal, a sequence to be predicted is generated; Initial features are extracted from the sequence to be predicted to obtain initial features, wherein the initial features represent mixed features corresponding to the diastolic blood pressure and the systolic blood pressure; Multi-resolution feature extraction is performed on the initial features to obtain the first feature and the second feature.

10. The apparatus according to claim 1, characterized in that, The acquisition module is also used for: Acquire multiple initial electromagnetic echo signals of the target object, wherein the multiple initial electromagnetic echo signals include an initial mixed signal of multiple chest initial electromagnetic echo signals reflected by the chest region of the target object and multiple neck initial electromagnetic echo signals reflected by the neck region; The multiple initial electromagnetic echo signals are subjected to arctangent demodulation processing to obtain multiple phase signals; The multiple phase signals are subjected to second-order differential filtering to obtain the multiple cardiac pulse signals.

11. The apparatus according to claim 1, characterized in that, The continuous blood pressure prediction generation model is trained using the following methods: Obtain the initial model to be trained and the sample training dataset, wherein the sample training dataset includes multiple consecutive blood pressure samples, multiple training physiological feature information, multiple training chest signals and multiple training neck signals; The multiple training physiological feature information, the multiple training chest signals, and the multiple training neck signals are input into the initial model to be trained for sequence generation processing to obtain multiple training sequences; The feature extraction is performed on the multiple training sequences to obtain multiple first training features and multiple second training features, wherein the first training features represent four-dimensional training features corresponding to systolic blood pressure, and the second training features represent eight-dimensional training features corresponding to diastolic blood pressure. The plurality of first training features and the plurality of second training features are input into the encoder for encoding processing to obtain a plurality of first training encoded features and a plurality of second training encoded features; The plurality of first training encoded features and the plurality of second training encoded features are subjected to feature fusion processing, and the resulting plurality of training fused features are input into the decoder for decoding processing to obtain a plurality of predicted continuous blood pressure; Based on the loss function, the training loss value is calculated according to the multiple predicted continuous blood pressure and the multiple continuous blood pressure samples; Based on the training loss value, the model parameters of the initial model to be trained are adjusted to obtain the trained continuous blood pressure prediction generation model.

Images